This paper presents our experience in developing a virtual assistant tailored for use in residential care facilities, aimed at reducing the workload of facility personnel by automating the collection of patient information and addressing routine inquiries. The system integrates a wide variety of AI methods, including natural language processing (NLP), computer vision, logical inference, ontologies, and knowledge graph technology for the system to interpret user messages and generate responses, ensuring contextual and personalized interactions. The knowledge graph is populated with semantic representations of user messages and serves as a verified source of factual knowledge, enhancing the system’s ability to provide accurate and reliable answers. Challenges in NLP and the integration of diverse subsystems were addressed through a hybrid approach that utilizes a specific language for unified knowledge representation, which improves explainability and addresses limitations inherent in large language models.